Motion Captioning, Motion Prior, Asset QA, Audio QA, OCT EYE, Reels,1P Search, Trigger, Trigger Eval, EGO-HOW-TO, 3DFY PROJECT
Motion captioning involves generating textual descriptions for complex motions, with temporal boundaries. Here's an overview of the scope: Motion Captioning Tasks: - Data Annotation: Labeling video or motion data with descriptions of actions, activities, or events. - Temporal Localization: Identifying the start and end times of specific actions or events within a motion sequence. - Caption Generation: Automatically generating captions for motion data, such as videos or 3D motion sequences {{IE_0}}¹{{/IE_0}}. Project Size: - CompMo Dataset: A large-scale dataset for dense motion captioning, featuring 60,000 motion clips with detailed captions and temporal boundaries. - Human Motion Insight (HMI) Dataset: A dataset comprising 115K video-description pairs focused on human movement, along with 1031K QA pairs {{IE_0}}¹{{/IE_0}} {{IE_1}}²{{/IE_1}}. Quality Measures: - Evaluation Metrics: Standard captioning metrics, such as BLEU-1/2/3/4, METEOR, ROUGE-1/2/L, and CIDEr, are used to assess.